COLA-Net: Collaborative Attention Network for Image Restoration

Local and non-local attention-based methods have been well studied in various image restoration tasks while leading to promising performance. However, most of the existing solely focus on one type attention mechanism (local or non-local). Furthermore, by exploiting self-similarity natural images, pixel-wise operations tend give rise deviations process characterizing long-range dependence due degeneration. To overcome these problems, this paper we propose a novel collaborative network (COLA-Net) for restoration, as first attempt combine local mechanisms restore content areas with complex textures highly repetitive details respectively. In addition, an effective robust patch-wise model is developed capture feature correspondences through 3D patches. Extensive experiments synthetic denoising, real denoising compression artifact reduction demonstrate that our proposed COLA-Net able achieve state-of-the-art performance both peak signal-to-noise ratio visual perception, maintaining attractive computational complexity. The source code available https://github.com/MC-E/COLA-Net .